Hi John,

Sorry about your camping weather - but I'm guessing I wouldn't have had such a fast and detailed reply if you'd been out in the sun fishing/hiking!

Well, that's worrying. I'll repeat my tests shortly but at the moment it seems a mystery why it was so glacial here.

Just to be absolutely sure it's the same collage, 4 > 01 has this preview, duration 22.04 s:

Hi Terry,

Oh no! That is a different 4 obj 01. Check that the preview window at the top says blaettern.wmv. I am doing the first export of it now. You are right, it will take a long time; now saying more than 20 minutes and climbing.

The one that I did before has 16:9 under it. There are several 4 object 1's. Why?

And, I can now confirm the same thing for VPX. This collage is a disaster.

That's a relief - glad I asked!

I ran it again using full duration of 22 s. Took 15 mins 17 s! (916 s, i.e. ratio of 42 to 1.)

Had the side of my case off to check for dust, thinking possible overheating, but it was surprisingly clean. And I gather 58 C (reported by CoreTemP) is not unreasonable for intensive processing (mine showing around 50%).

I was part way through my second repeat test with different settings when I checked the thread and saw your welcome reply. Now cancelled the render.

It is admittedly a relatively complex collage, but still seems incredibly slow. Will never use it again

Not sure why you have two identically labelled collages for 4 landscape objects. What preview file is associated with the one you tested before?

John/Peter/Jeff/anyone - any thoughts?

The first one that I used was: lsplsls_left_third1

The labels are slightly different. It seems that the sort is throwing me off and for some reason the icons don't show up in VPX.

I should have looked more closely at the heading of this post that has > rather than ^ .

Thanks John.

I may test a few others in the 4> category some time.

It's incredible how long MEPP 2016 (and I'm guessing 2017/2018?) take to render apparently simple collages.

I've just included the two landscape object > 02 04 in my current (1920x1080 50 fps) project. The collage, applied to two 4 second JPGs, is just under 13 seconds long and took nearly 8 minutes to render. (The only reason I persisted was because I want to halve its duration; I find most of the built-in collages too long.) A small number of those (and presumably others in the Magix repertoire) included unthinkingly in a short project would have me very impatient about total render time!

When I get a chance I'm going to fire up my much slower XP PC on which I have some old versions of MEP installed and test it for comparison. But I'm fairly sure I'd have noticed this issue, even though until recently I've always worked at 25 fps.

Hi Terry

. . . . the two landscape object > 02 04 in my current (1920x1080 50 fps) project. The collage, applied to two 4 second JPGs, is just under 13 seconds long and took nearly 8 minutes to render. . . . .

I get the following results using the above on 2 images, original size 3840 x 2160:

Exporting to MXV and MP same settings as your project

• MEP 2016
  
  MXV 7m 00s - no HWA (as expected)
  MP4 7m 02s - used HWA
   
• MEP 2018
  
  MXV 4m 17s - no HWA (as expected)
  MP4 4m 20s - used HWA

I checked the MEP 2018 export twice for each format due to the significant difference in times compared to 2016.

HTH

John EB

Hi John,

Thanks for that thorough test. Although 2018 is much better, don't you think a ratio of 20:1 still seems remarkably high for two objects and a smattering of KFs?

What's the oldest working version of MEP you have installed?

Hi,

Just to confirm, it is "freeze2,wmv" that shows up as the source?

With VPX on my Surface Pro X, no HWA, 5:51 to export to MXV. It's like watching paint dry, or in my case, watching the varnish on our newly sanded and stained wood floors dry whilst camped out in the neighbour's yard.

John CB

Hi Terry

Out of interest I also ran the same project through VPX and got the following:

Intel (default) codec

MP4 4m 13s (using HWA)

MXV 4m 13s (no HWA as expected

MainConcept codec

MP4 4m 38s - no HWA - however, the resulting video was poor quality with visible softness and lower contrast.

. . . . What's the oldest working version of MEP you have installed? . . .

MEP MX is the oldest working - Windows 10 Creators Edition (1703) terminated MEP 17.

In MX the Collage [>] 02 Obj 04 is not the same as later versions.

The project from MEPP did load with the correct effects and took 1m 35s to export, however, the MainConcept is the only one available and as for VPX the quality was poor with visible softness and lower contrast.

. . . . Although 2018 is much better, don't you think a ratio of 20:1 still seems remarkably high for two objects and a smattering of KFs? . . . .

IMHO what is actually happening in the collage is the determining factor. If you look at the effect there is a lot of image recalculation being done as the images are rotated, moved and zoomed.

HTH

John EB

Hi John CB

. . . . Just to confirm, it is "freeze2,wmv" that shows up as the source? . . . .

In MEP the demo file name for the effect is not displayed, however that is the correct one.

. . . . It's like watching paint dry, or in my case, watching the varnish on our newly sanded and stained wood floors dry whilst camped out in the neighbour's yard . . . .

Hope the floors dry quicker than that, winter is just round the corner.

HTH

John EB

Hi John EB,

It's worse than that. Second round is Monday when they come to do the kitchen and dining room. I'm leaving for Africa again on Saturday, Sunday in London, for a week, so I have to get the furniture moved back into the living room, empty and prepare the 2 rooms, between Friday night and Saturday afternoon. My wife is not amused; she is stuck alone in the trailer all of next week.

I should have filmed this event.

In VPX when running the preview of a collage or anything else, the file name is shown at the top of the source monitor; one more little advantages of VPX, I guess.

John CB

Hi,

Just to confirm, it is "freeze2,wmv" that shows up as the source?

With VPX on my Surface Pro X, no HWA, 5:51 to export to MXV. It's like watching paint dry, or in my case, watching the varnish on our newly sanded and stained wood floors dry whilst camped out in the neighbour's yard.

John CB

Hi both,

Yes. (Quick way to check is to right click the collage icon and choose Properties.)

Thanks for the further feedback. It prompted me to repeat it in a fresh MEPP 2016 project (HW on) and this time I got 4:12. I now reckon that yesterday a rare Win 10 image backup was running in the background (or possibly an automatic Defender scan that arose when the backup found a (false positive) trojan.

Still, although I now have slightly quicker drying paint, and accept that some fairly complex stuff happens, that collage and several of its immediate relatives are going on my list of 'Collages to avoid'.

Camping in a tent at night? Hope your tent zipper is a strong one?

Hi Terry

Love the cartoon. Not so funny for John CB though!

I ran the 13 second 02 04 collage on 2 x 4592x3064 jpegs in VPX8 and got the following:-

Intel (default) codec

MP4 2m 40s (using HWA) - GPU-Z monitor shows virtually no activity on either of my GPU's

MP4 17sec (using HWA + Calc.FX on GPU) - GPU-Z shows 35-45% load on Nvidia GPU and 13% on iGPU

but

MP4 3m 03 (using HWA + Calc.FX on GPU and Anti Interlace Filter) - surprisingly GPU-Z shows only 5-12% on Nvidia and 2% on iGPU

MainConcept codec

MP4 2m 46s - no HWA (ie.SW) - however, the resulting video was poor quality with visible softness and lower contrast. NB I have reported several times in the forum that the MC SW encode is inferior to the MC HW .m2ts export and Intel HW and SW MP4 export.

So:-

As John EB states

"IMHO what is actually happening in the collage is the determining factor. If you look at the effect there is a lot of image recalculation being done as the images are rotated, moved and zoomed."

Playing around in the FX Keyframer and isolating various parameters whilst monitoring GPU load - the main load on the GPU seems to be from the rotation which occurs on 2 separate clips simultaneously ie the foreground and faint background. This effect is mainly accelerated on the discreet Nvidia GPU so without the Calc.FX on GPU ticked there is little HWA occuring on either GPU (at least in my case and probably in cases where you only have an iGPU).

The strange one is the 3m 03secs Intel HW export with the FXonGPU and Anti Interlace filter ticked. The Anti interlace filter seems to have killed most of the Rotation effect processing on the Nvidia GPU. This is not altogether surprising, for one because the AI filter is actually a softening filter which is processing intensive and probably not GPU accelerated, which loads the already saturated CPU further. Not really knowing entirely how this all interacts on the hardware, my belief is that this causes a bottle neck for getting the info to the GPU and the HWA falls over.

Interesting that the AI filter is on by default on JPEGs in the Rt click menu and Key framer although theoretically not activated until the box is ticked in the Export dialogue, but I have my suspicions that one setting is affecting the other in some other way.

My other theory is that encoding/decoding during processing of an MPEG4 to MPEG4 is very much accelerated by the Quick Sync block on the Intel (i)GPU but you will see less accel. on a JPEG to MPEG4 encode because it can not benefit on the decode side of the processing, or if in fact at all as witnessed by the 1-2% load experienced by my Intel HW encode @ 2m 40s.

@ Johnbaker

I checked the MEP 2018 export twice for each format due to the significant difference in times compared to 2016.

They did claim 2018 is 5x faster........haha!😄

Peter

Hi Peter,

Wow, that's a revelation! I'd never regarded 'Apply anti-interlace filter (suppresses fliickering)' as a significant factor affecting render time, but I sure am glad your tests revealed that.

Here are my results just now for all eight combination of those three settings:

BTW, you'll note that my YYY result above is similar to my earlier one. I have to conclude the 4:12 was a mistake by me.

All my results use the built in Intel graphics.

P.S: nVidia apart, are you feeding your PC with any particular steroid you can share with us?

Terry, East Grinstead, UK

Hi

Interesting - I get the following - same project as before

MP4 export as project.

Calculate video effects on GPU Off

• HWA with interlace filter 4m 15s - previously 4m 13s
• HWA no interlace filter 4m 10s - turned off in program settings and export dialog

Very little difference to the time with the filter on 😟

Calculate video effects on GPU On

• HWA with interlace filter 4m 08s
• HWA no interlace filter 1m 18s

Significant difference with no interlace filter - this would confirm Peters suggestion about the AI not being accelerated.

Previously, in earlier versions of MEP, I have avoided the use of the Calculate video effects on GPU option on the iGPU because of it causing issues - looks like it is fixed.

. . . . They did claim 2018 is 5x faster........haha! . . . .

Well it looks like the claim is valid - 4x faster in my case using the HD4600 GPU is a bonus, perhaps when it comes to upgrading my PC to an i5 or i7 gen 8 processor I will get that missing 25% increase.😁

Cheers

John EB

Hi

Good to see you both have been able to get some better times on the encodes.

I'd never regarded 'Apply anti-interlace filter (suppresses fliickering)' as a significant factor affecting render time

Further....I believe the AI filter is a pre-processing filter that adds a median blur in front of the encoder similar to a feature in the encoder that old versions of MEP used to use prior to Magix getting into bed with Mainconcept. The idea there was to increase encoding efficiency by blurring the noise prior to the MPEG compression but now the idea is to reduce detail by blurring to suppress aliasing when scaling hires material to low res output. The blurring is quite noticeable in my opinion on HD encodes so I would not use it, nor is it necessary except perhaps on SD encodes of hires material.

P.S: nVidia apart, are you feeding your PC with any particular steroid you can share with us?

Well Terry, your test No.7 result beats mine.....but I think sharing the load between a discreet card and the iGPU helps in encodes and preview, plus I have overclocked my CPU to 4.6GHz using the automatic overclock function available through the ASUS Ai Suite 3 utility, which definitely helps.

John EB - you mentioned you disabled the AI filter in the program settings - what setting is that?

Best

Peter

Sorry I didn't have time to try test the collage on my desktop with MEP2018 Plus installed. I'll try in a week or so when I get home. I really want to see if there is a difference in speed between machines and VPX and MEP2018.

Terry, there are no bears in the neighbourhood, only raccoons, squirrels and skunks. Bernadette keeps the first two away and, fortunately, I have managed to keep her away from the skunks that have been digging for grubs in our front lawn. My wife has to deal with the camping, the dog, and the next round of floor finishing this week while I travel. I'm currently camped out at Heathrow between flights, heading for one of those bed chairs that they have in the lounge. The night was too short. And no bears here.

My best,

John CB

Hi Peter

. . . . AI filter in the program settings . . . .

This one which is actually for video, however I tried it to see if it had any effect on still images, does not look like it does.

As you can see in the dialog, it is not clear whether this applies to videos or images or both!

John EB